FLUORIDATION TRIAL CONTROLS: ERRORS, OMISSIONS AND MIS-STATEMENTS: INTRODUCTION
The fluoridation trials that were conducted in the cities of Grand Rapids, Newburgh and Evanston, in the U.S.A., and the two independent ones in Brantford, Canada, are of more than ordinary importance, because they constitute the main experimental evidence which has led to the introduction of this process as a public health measure. The fluoridation hypothesis is "that a concentration of about I part per million of fluoride in the drinking water, mechanically added, inhibits the development of dental caries in the teeth of the users of the water" (Brown, McLaren and Stewart, 1954b). In 1956 Nesin pointed out: "It must be emphasized that the fluoridation hypothesis in its entirety rests on a very narrow base of selected experimental information. It is this very base which is vulnerable to scientific criticism. And, it is upon this very narrow base that the very impressive array of endorsement rests like an inverted pyramid."
The safety of artificial fluoridation has been questioned by a number of eminent authorities such as Hicks (1956) and Sinclair and Wilson (1955). In 1955 Box stated: "It is my considered opinion that the artificial fluoridation of water supplies, on a wholesale basis, should not be advocated or adopted until fully sufficient findings show that there are no harmful sequelae from a gingival or periodontal standpoint."
However, these questions need be considered only if the overall dental benefits of fluoridation are demonstrated beyond reasonable doubt, and are also found to be worthwhile from a socio-economic point of view. No suggestion has been made that fluoridation has other than dental benefits.
It has been widely accepted that the existence of marked dental benefits has been established, and the literature abounds with references to reductions of about 60 per cent in dental caries as a result of fluoridation. However, the published works contain little consideration of the numerical data reported from these trials, as distinct from mere statements of percentage reductions in the caries attack rates.
A preliminary examination revealed that reports of these studies contain errors and show omissions, and statements made in regard to results are not justified by published data; therefore further study has been made of these crucial trials. This study attempts to evaluate their controls, and the discussion is limited to examination of published reports of (i) method of selection of control cities; (ii) their suitability; (iii) the experimental and statistical processes used in gathering and analysing the data (iv) the results stating the dental caries attack rates; (v) some comments made by the authors of these trials (and by others) on these results.
The aim will be to investigate the reliability of the results reported, to assess the adequacy of the controls that were set up and to evaluate the accuracy of the statements made concerning the data obtained.
Before discussing the procedure adopted in each of these studies, several basic matters that are of importance in a fluoridation trial will be considered.
The necessity for controls. Blayney and Tucker (1948) were correct in stating that "A study of this nature must have an adequate control." The necessity for such a procedure was recognized by the authors of four out of five of these studies. Cities with "fluoride-free" water supplies were selected as controls, and comparisons were made with towns which possessed water supplies with a fluoride content obtained from natural sources, which approximated the concentration which has been called the "optimum" one (Dean, Arnold, Jay and Knutson, 1950; Brown, 1951; Ast and Chase, 1953; Hill, et al., 195 1). It is to be noted that in the trial conducted in Brantford by the City Health Department (Hutton et al., 1951) no provision for controls was made.
Requirements of a control. In an experiment such as the fluoridation of the water supply of a city, whereby the whole of its population is subjected to treatment (fluoridation), it is necessary to obtain the control data from subjects who live in a city or cities with "fluoride-free" water supplies. In determining the cities which are to participate in the trial, in order to increase the sensitiveness of the experiment, it is advantageous to employ ones which are alike in as many respects as it is practically convenient to consider. Of course, as Fisher (1951) pointed out, "the uncontrolled causes which may influence the result are always strictly innumerable."
Because of the nature of these experiments, three main points of similarity must be considered and described. These are (a) the water supply; (b) the climate; and (c) the dental caries attack rates. Other factors, such as socioeconomic status, are of less importance; their influence may be reflected in the caries attack rates.
(a) In its statement of its official policy on this matter, the American Water Works Association (1949) said that the experimental verification of the fluoride-dental caries hypothesis "obviously necessitates the use of a nearby control city with a water supply comparable in all respects to that to which fluoride is being added." The Association referred to "the possible influence, on the fluoride potency, of other chemical constituents of natural waters, insofar as these and other variables may affect the action of fluoride on the control of caries in a human population." In 1942 Deatherage reported that "It is these soft waters which cause the most severe mottled enamel." Therefore, the fact that both the test and the control city in a fluoridation trial obtain their water from the same source does not remove the necessity for a study of the composition of the water. Dean, Jay, Arnold, McClure and Elvove (1939) recognized this, stating, "the possibility that the composition of the water in other respects may also be a factor should not be overlooked. For this reason it seems highly desirable that dental caries studies should be accompanied by complete chemical analyses of the dam waters, including a search for the comparatively rare elements." However, in none of these trials was the composition of the water stated.
(b) The climate of a city is an important factor in determining the average amount of salts ingested from the water supply, because of its influence on the volume of water consumed by humans. Therefore, cities that are to be compared should not only have water supplies that have a closely comparable composition, but the climates of the cities should also be very similar.
(c) As the main aim of fluoridation is to reduce the dental caries attack rates, it is obviously of importance that the cities to be compared should have closely comparable dental caries rates within yearly age groups, of children. This information can be obtained only by conducting at least one survey in the cities that are suitable for comparison on other grounds, so that the fact that the caries attacks rates are similar is established prior to the fluoridation of the water supply of one of them.
Random sampling. The fundamental importance of random sampling has been acknowledged for many years. In designing an experiment, as Quenouille (1952) said, "it is necessary to allot the treatments to the available material at random if unbiased estimates of both the effect of the treatments and also the reproducibility of the effects are to be obtained." Therefore, a random device should be employed to determine which of the participating cities is to be the test one.
Examiner variability. In experiments in which, of necessity, the subjective judgment of examiners is employed, an important consideration is the assessment of "between-examiner" and "within-examiner" variability. The former type of variability is disclosed when different examiners observe the same subjects, and the latter type is seen in the different results reported by the same examiner inspecting the same subjects on different occasions, but which are sufficiently close together to ensure that the dental condition has not undergone appreciable change.
The important effect which examiner variability can have on the results of a study of dental caries attack rates was pointed out by Radusch (1941) and by Dunning (1950). A recent example is seen in the paper of McCauley and Frazier (1957). Their Table I shows that in the examinations made by one examiner in 1955 of Negro boys and girls who were six years old, in both sexes the DMF rate per 100 teeth erupted, and also the DMF rate per child, were found to be about four times as great as those reported for the same age groups in 1952 when they were examined by several examiners. The authors considered that "it is entirely possible that the 1952 findings were influenced by a bias stemming from subjective differences in the appraisal of tooth decay by different dentists". Between-examiner variability of such a magnitude can, of course, vitiate the results of a study. Unless the examiner variability is determined, and is taken into account, the conclusions drawn from a study of caries attack rates must be treated with reserve.
Examiner bias. In designing an experiment of this nature, one aim should be to eliminate examiner bias. This may arise if the examiners know whether the children they are examining belong to the test or to the control city. One method of doing this is to transport to a common examination centre the small number of children, some from the test and some from the control city, that can be examined each day; the examinations being conducted in a random order which is unknown to the examiners. It is not suggested that in the absence of such precautions the examiners exhibited intentional bias; indeed, as Armitage (1954) pointed out, "through fear of being biased" the judgment of an examiner may be influenced.